Highlight-Based Color Corrections
Here is this a photo of Ellie Guttman of Washington DC,
out enjoying a day on the water in Alaska. A blue color cast (common
in outdoor scenes) pervades the photo. We know there is a blue color
cast by measuring the RGB values in the what is supposed to be the
diffuse white highlight area found in Ellie's cap.
 |
Click the image
to download a TIFF to work with. |
A neutral white
highlight should have equal RGB (red, green, and blue) values. Here
you can see that the RGB point #1 values are 235, 238, and 241 respectively.
The blue value (241) is significantly higher than the red and green
values of 235 and 238. This is proof of the blue color cast.

Using
the Curves dialog and the 235 red value as the target highlight value,
we lower the blue and green highlight values until they are
235, and then make a similar adjustment in the midtone.

Step-by-Step
1. Open and duplicate
the Ellie Boot_Start.tif image that you downloaded above. Name the
duplicate Ellie Boot_Finish.
2. Activate
your Color Sampler tool (found under the Eyedropper tool in the Tools
palette or by pressing Shift+I). This tool will allow you to record
and monitor
up to
four
RGB value
sets.
Also activate the Info palette (Window > Info or F8) and set the
Color Sampler tool's Sample Size to 3 x 3 Average.
3. Locate
and click, with the Color Sampler tool, on the white neutral diffuse
highlight
in Ellie's cap. A Color Sampler point is placed where you click on
Ellie's cap. The RGB values for this point are recorded as point
#1 and displayed at the bottom of your Info palette. Your RGB value
set should be close to the one I measured above (about 235, 238,
241).
4. Activate
your Curves dialog (Image > Adjustments > Curves or
Ctrl/Command+M).
5. Activate
the Blue channel curve form the Curves Channel menu.
6. Click
on and pull the highlight end of the Curve down until the blue value
for point
#1 in your Info palette equals the Red value (here 235). Then click
and drag the midpoint of the Blue channel down an amount equal to
the adjustment you made on the highlight point.
Note: This midtone
adjustment is made assuming that the blue color cast you identified
in the highlight is pervasive throughout the image's tonal range.
While this is not always the case, it is a shortcut assumption that is often
useful.
7. Now,
activate the Green channel curve, and perform a similar adjustment
to this channel.
Click on and pull the highlight end of the Curve down until the Green
value for point #1 in your Info palette equals the Red value (235).
Then click and drag the midpoint of the Green channel down an amount
equal to the adjustment you made on the highlight point.

8. Click the OK button to apply these Curves
changes. Now compare your Ellie Boot Start and Finish images.

The results of the highlight-based correction are:
 |
|
 |
| |
1. The neutral white highlight pops.
2. The yellow boot has better saturation.
3. Ellie's skin looks healthier.
4. The original subtle green in Ellie's coat shows through.
5. There is an overall reduction in the blue cast across the
image, resulting in overall improvement of the image ... even
the spruce trees in the background look better (more saturated). |
|
 |
|
 |
We didn't make a huge color correction here, but you
can see that the results are a noticeable improvement.
Skin Tone Adjustments
Another common, and critical type
of color correction is the RGB ratios in skin tones.
Here we have Nancy
Lasater, Ellie's mom, also out on the water in Alaska.
 |
Click the image
to download a TIFF to work with. |
The RGB values
of her face show that blue (206) is above the green (202).
This violates the general facial skin tone rule that R > G > B
(in humans anyway, Klingons are another matter). Using the Curves dialog,
we lower the overall blue values until Nancy's skin tones values conform
with the R > G > B rule.
Step-by-Step
1. Open and duplicate the
Nancy_Start.tif image that you downloaded above. Name the duplicate
Nancy_Finish.
2. Activate your
Color Sampler tool and Info palette, and set the Sample Size to
3 x 3 Average as you did for the previous photo.
3. Locate and click on
the a well-lit portion of Nancy's face (I used a place on Nancy's chin
above) with the
Color Sampler tool. A Color Sampler point
will be placed where you click on Nancy's face. The RGB values for
this point will be recorded as point #1 and displayed at the bottom
of the Info palette. Your RGB value set should be close to the one
measured above (about 237, 202, 206). Note the Green is well below
the Red (as it should be if R > G > B). The blue is the problematic
value, so you will want to lower the blue value.
Note: For the purposes
of brevity, you are using only one Color Sampler point. However, you
may want to locate two or three points (I often
do) to make sure that you are making an overall correct adjustment.
Creating multiple Color Sampler points will create multiple Info
Palette monitoring points as well. One of the advantages of using the
Color
Sampler tool is that you can measure, record, and monitor values
of up to four separate points.
4. Activate
your Curves dialog (Image > Adjustments > Curves or
Ctrl/Command+M).
5. Activate
the Blue channel curve form the Curves Channel menu.
6. Press your Ctrl (PC)
or Command (Mac) key and click on top of the Color Sampler point you
placed on Nancy's face.
This will create
a
control point on the
Blue Curve at the point in the tonal range where the Color Sampler
point is placed.
Note: If you are using
multiple Color Sampler points, you can create and edit several controls
points. But first
try editing
just one
point while monitoring the RGB values on the other Color Sampler
points.
7. Now click
on the blue channel control point on the curve and drag it down while
you monitor the Blue value in displayed in Color Sampler
point #1. Lower the Blue curve until the Blue value is well separated
from
the Green value.

Note: How much separation
you create will depend upon the image and the current Red and Green
values. A good
starting
place on
many images
is to create a 5:4:3 ratio between Red, Green, and Blue values
(in well-lit areas of the image). If this seems too much separation,
try
a 7:6:5 ratio.
In this image, since Nancy has red skin and hair, I have allowed
the red value to be more separated from the green and blue
values than
in other images. You may want to place and monitor a Color
Sampler point in a neutral area of your image, such as the gray under
jacket I have sampled here, to help you determine how much
you
adjust
the channels.

8. Click the OK button to apply your Curves
changes. Now compare your Nancy Start and Finish images.

The results of this color correction are:
 |
|
 |
| |
1. Nancy's skin tone looks more natural. She looks less like
she is turning blue from lack of oxygen.
2. Nancy's gorgeous red hair really pops now once all that
excess blue is removed!
3. The orange in Nancy's jacket has better saturation. |
|
 |
|
 |
Both of the above scenarios are examples of honest-to-goodness
color corrections. The color values were wrong, and when they were
corrected, the images showed noticeable improvements.
Color Adjustments
As we mentioned earlier, not all color
changes represent color corrections. You may want to change the color
in an image to change its mood
or its emphasis, with neither the original or the final image being
correct, just different from each other.
Here
is a photo I shot of the southern Kenai Mountains in Alaska on an early
morning kayak trip.
 |
Click the image
to download a TIFF to work with. |
The early morning light was quite yellow and projected
a very calm mood. This image is just fine the way it is; there is nothing
wrong with
the color. But we can change its mood by reducing
the yellow cast and creating a more neutral or even a bluish cast.
This can be accomplished by equalizing or neutralizing
the data in histograms using the Levels dialog in Photoshop.

Step-by-Step
1. Open and duplicate the Kenai_Start.tif image you downloaded
above. Name the duplicate Kenai_Neutralize.
2. Activate your Levels
dialog (Image > Adjustments > Levels or
Ctrl/Command+L).
3. In the default RGB Channel view, click and drag the
RGB shadow slider from its far left position to the right into the
beginning
of the shadow
data of the image.

Note: This will darken the shadow areas of the
image, thereby increasing the overall contrast of the image. If you
want to make the darkest
area solid black (with no detail), drag the shadow slider further
into the shadow data portion of the histogram.
4. View the histogram
data of all three individual channels. Note how the starting Blue
channel highlight is offset to the left of
the Red
and Green channels' highlight data. This is the source of the
yellow (red + green) color cast.
5. Select the Blue channel from the
Levels Channel menu.
6. Drag the highlight slider of the Blue channel
to the left, until it is just under the beginning of the highlight
end of
the histogram
data.

Note: This will place the beginning
of the Blue channel histogram highlight data in the same location as
the Red and Green
channels, thereby neutralizing
the highlight data and removing much of the yellow (red
+ green) color cast from the image.
7. Click the OK button to apply
your Levels changes. Now compare your Kenai Start and Neutralize
images.
We can further enhance the blue shift using Levels or
Curves. Using Curves is my preferred method.
1. Duplicate the Kenai_Neutralize image
you just created above. Name the duplicate Kenai_Blue.
2. Activate
your Curves dialog (Image > Adjustments > Curves or
Ctrl/Command+M).
3. Activate
the Blue channel curve form the Curves Channel menu
4. Now, click on the
midpoint of the Blue channel curve and drag it up 10%.

5. Click the
OK button to apply your Curves adjustment. Now compare your Kenai
Start, Neutralized, and Blue images.

Note how this last Curves adjustment
enhances the blue value in both the sky and water. Also be aware
that you can create selections
of
specific portions of an image, such as the sky and/or the water
here, and adjust the color values for only that/those portion(s)
of the image.

These two color adjustments do not represent corrections,
as there was nothing wrong with the images. But they certainly represent
significant
color adjustments.
Color Correction and Adjustment Tips
Deciding
on Color Changes
A good approach to making color changes
is to ask yourself the following questions and apply these actions.
 |
|
 |
| |
1) Is there any color correction that needs to be applied
to this image? If so determine (using your Info palette and
histograms) what those corrections are.
2) Make any needed color
corrections prior to proceeding with any color adjustments.
3)
What, if any, non-correction-based color adjustments might
you like to make? Decide which tools and methods you
should use to accomplish those
corrections. |
|
 |
|
 |
Tip: Always apply the change adjustments
to copies of your original
images so you can always return to the original. This
is especially true if the images
belong to someone else!
Taz's Top Five Color Correction
Tools
 |
|
 |
| |
1) The Info palette provides me foolproof RGB and CMYK image
values ... the
Info palette will not lie!
2) The histogram in the Levels dialog shows
me where the data is and how it is distributed.
3) The highlight
and shadow sliders in the Levels dialog allow me to adjust
the highlight and shadow points while viewing
an image's histogram data.
4) The Curve dialog allows me to
make specific image corrections anywhere within the tonal range
of an image.
5) The highlight and midtone Eyedropper tools in both
Levels and Curves dialogs allow me to make fast and accurate
preset highlight
and shadow adjustments to images,
especially when I am working with target-based color corrections. |
|
 |
|
 |
back to top |